Neura Intelligence Index
Category Comparison
Benchmark-by-Benchmark
reasoning
| Benchmark | Gemini 2.5 Pro | Gemini 3.5 Flash |
|---|---|---|
| GPQA Diamond | 83 | — |
| Humanity's Last Exam | 17.8 | 40.2 |
| ARC-AGI v2 | 4.9 | 72.1 |
math
| Benchmark | Gemini 2.5 Pro | Gemini 3.5 Flash |
|---|---|---|
| AIME 2025 | 83 | — |
coding
| Benchmark | Gemini 2.5 Pro | Gemini 3.5 Flash |
|---|---|---|
| SWE-bench Verified | 63.2 | — |
| SWE-Bench Pro | — | 55.1 |
multimodal
| Benchmark | Gemini 2.5 Pro | Gemini 3.5 Flash |
|---|---|---|
| MMMU | 79.6 | — |
| MMMU-Pro | — | 83.6 |
| CharXiv Reasoning | — | 84.2 |
agent
| Benchmark | Gemini 2.5 Pro | Gemini 3.5 Flash |
|---|---|---|
| Toolathlon | — | 56.5 |
| MCP Atlas | — | 83.6 |
Category Winners
coding
44.6
math
57.4
reasoning
74.6
general
69.1
multimodal
80.4
agent
72.3